COS 161-10 - Exploring invasibility with species distribution modeling: How does fire promote cheatgrass invasion within lower montane forests?

Thursday, August 10, 2017: 4:40 PM
B116, Oregon Convention Center
Jamie L. Peeler, Geography Department, The Pennsylvania State University, University Park, PA and Erica A.H. Smithwick, Geography Department and Intercollege Graduate Degree Program in Ecology, The Pennsylvania State University, University Park, PA

Cheatgrass (Bromus tectorum) is an invasive winter annual that has spread rampantly since its introduction from Eurasia in the late 19th century. The cheatgrass-fire cycle has been well described for sagebrush steppe ecosystems in the Great Basin. However, mechanics of this cycle are not well understood for forest ecosystems in the western United States. Our study explored how fire enhanced invasibility within a recently burned lower montane forest in the Greater Yellowstone Ecosystem (GYE). We addressed three questions: (Q1) Does fire alter the potential distribution of cheatgrass? (Q2) What site attributes are most strongly associated with cheatgrass presence and do these differ between fire and no fire scenarios? (Q3) Does fire alter invasion pathways and enhance potential spread? To do so we created spatially explicit Random Forests models with the ModelMap package in R. The first model excluded plots within the fire perimeter to generate a neutral model (n = 71), while the second model included all plots sampled (n = 93). Each model produced a probability surface on cheatgrass presence across our study site in the GYE. Additionally, we modeled invasibility in our lower montane forest using Circuitscape.


Our neutral and fire models performed well with an Area Under the Curve (AUC) of 0.94 and 0.92 respectively. We observed terrain and climate factors as top predictor variables for the neutral model: precipitation, elevation, aspect, minimum temperature in January, and maximum temperature in July. In contrast, a slightly different suite of predictor variables controlled the fire model: canopy cover, precipitation, elevation, aspect, and minimum temperature in January. The mean decrease in accuracy for canopy cover increased from 9.9% to 21.5%. A threshold of approximately 30% canopy cover was also observed for predicting cheatgrass presence/absence. Circuitscape results showed that potential spread increased in areas where fire reduced canopy cover within our study site in the GYE. Accordingly, fire appears to increase invasibility through the reduction of canopy cover in climatically suitable forests – an observation that could have implications for preventing a cheatgrass-fire cycle that locks post-fire landscapes into an alternative stable state in the western United States.